Search Images Maps Play YouTube News Gmail Drive More »
Sign in
Screen reader users: click this link for accessible mode. Accessible mode has the same essential features but works better with your reader.

Patents

  1. Advanced Patent Search
Publication numberUS5893092 A
Publication typeGrant
Application numberUS 08/880,807
Publication dateApr 6, 1999
Filing dateJun 23, 1997
Priority dateDec 6, 1994
Fee statusPaid
Also published asUS5642502, US6088692
Publication number08880807, 880807, US 5893092 A, US 5893092A, US-A-5893092, US5893092 A, US5893092A
InventorsJames R. Driscoll
Original AssigneeUniversity Of Central Florida
Export CitationBiBTeX, EndNote, RefMan
External Links: USPTO, USPTO Assignment, Espacenet
Relevancy ranking using statistical ranking, semantics, relevancy feedback and small pieces of text
US 5893092 A
Abstract
Search system and method for retrieving relevant documents from a text data base collection comprised of patents, medical and legal documents, journals, news stories and the like. Each small piece of text within the documents such as a sentence, phrase and semantic unit in the data base is treated as a document. Natural language queries are used to search for relevant documents from the data base. A first search query creates a selected group of documents. Each word in both the search query and in the documents are given weighted values. Combining the weighted values creates similarity values for each document which are then ranked according to their relevant importance to the search query. A user reading and passing through this ranked list checks off which documents are relevant or not. Then the system automatically causes the original search query to be updated into a second search query which can include the same words, less words or different words than the first search query. Words in the second search query can have the same or different weights compared to the first search query. The system automatically searches the text data base and creates a second group of documents, which as a minimum does not include at least one of the documents found in the first group. The second group can also be comprised of additional documents not found in the first group. The ranking of documents in the second group is different than the first ranking such that the more relevant documents are found closer to the top of the list.
Images(19)
Previous page
Next page
Claims(10)
I claim:
1. A method for retrieving relevant text data from a text database collection in a computer without annotating, parsing or pruning the text database collection, comprising the steps of:
(a) searching a text database collection in a computer using a first search query of natural language to retrieve a first group of selected small pieces of text, where each of the selected small pieces of text corresponds to a document;
(b) ranking each of the selected small pieces of text into a first ranked list of relevant documents;
(c) applying feedback information based on a manual determination of the relevancy of each of the selected small pieces of text in the first ranked list to automatically create a second search query, the second search query being different than the first search query;
(d) repeating steps (a) to (b) to form a second ranked list, wherein the second ranked list includes a second group of selected small pieces of text, and the second group is different than the first group.
2. The method for retrieving relevant text data of claim 1, wherein each of the small pieces of text includes at least one of:
a sentence, a phrase, and a semantic unit.
3. The method for retrieving relevant text data of claim 1, wherein the second search query includes:
at least one less word from the first search query.
4. The method for retrieving relevant text data of claim 1, wherein the second search query includes:
at least one additional word to the first search query.
5. The method for retrieving relevant text data of claim 1, wherein the second group includes:
at least one less document that had been listed in the first group.
6. The method for retrieving relevant text data of claim 1, wherein the second group includes:
at least one additional document that was not found in the first group.
7. The method for retrieving relevant text data of claim 1, wherein the second ranked list includes:
a different ranked order of documents than the first ranked list.
8. A method for retrieving relevant text data from a text database collection in a computer without annotating, parsing or pruning, comprising the steps of:
(a) searching a text database collection in a computer using a first search query to retrieve a first group of selected small pieces of text, where each of the selected small pieces of text corresponds to a document;
(b) semantically weighting the selected small pieces of text to form document weighted values for each of the selected small pieces of text in the first group;
(c) semantically weighting the first search query to form query weighted values;
(d) combining the query weighted values and the document weighted values to form similarity values for each of the selected small pieces of text;
(e) ranking the similarity values for each of the selected small pieces of text to form a first ranked list;
(f) automatically updating the first search query into a second search query based on feedback information on whether documents in the first ranked list are relevant,
(g) repeating steps (a) to (e) to form a second ranked list, wherein the second ranked list includes a second group of selected small pieces of text which is different than the first group.
9. The method for retrieving relevant text data of claim 8, wherein each of the small pieces of text includes at least one of:
a sentence, a phrase, and a semantic unit.
10. A method for retrieving relevant text from a text database collection in a computer without annotating, parsing or pruning the text database collection, comprising the steps of:
(a) searching a text database collection in a computer using a first search query to retrieve a first group of selected text;
(b) ranking each of the selected text to form a first ranked list;
(c) determining relevancy of each of the selected text with a manual pass-through of the first ranked list; and
(d) automatically updating the first search query based on the relevancy determination of the manual pass-through into a second search query, the second search query being different than the first search query; and
(e) searching the text database collection using the second search query to retrieve a second group of selected text being different than the first group.
Description

This is a Divisional of application Ser. No. 08/350,334 filed Dec. 6, 1994 which issued as U.S. Pat. No. 5,642,502 on Jun. 24, 1997.

This invention relates to natural language data processing, and in particular to a method and system for the retrieval of natural language data. This invention is related to U.S. patent application Ser. No. 08/148,688 filed on Nov. 5, 1993, now U.S. Pat. No. 5,576,954, which is incorporated by reference. This invention was developed with grant funding provided in part by NASA KSC Cooperative Agreement NCC 10-003 Project 2, for use with: (1) NASA Kennedy Space Center Public Affairs; (2) NASA KSC Smart O & M Manuals on Compact Disk Project; and (3) NASA KSC Materials Science Laboratory.

BACKGROUND AND PRIOR ART

Locating information using large amounts of natural language documents (referred to often as text data) is an important problem. Current commercial text retrieval systems generally focus on the use of keywords to search for information. These systems typically use a Boolean combination of keywords supplied by the user to retrieve documents. See column 1 for example of U.S. Pat. No. 4,849,898, which is incorporated by reference. In general, the retrieved documents are not ranked in any order of importance, so every retrieved document must be examined by the user. This is a serious shortcoming when large collections of documents need to be searched. For example, some data base searchers start reviewing displayed documents by going through some fifty or more documents to find those most applicable.

Statistically based text retrieval systems generally rank retrieved documents according to their statistical similarity to a user's search request (referred to often as the query). Statistically based systems provide advantages over traditional Boolean retrieval methods, especially for users of such systems, mainly because they allow for natural language input.

A secondary problem exists with the Boolean systems since they require that the user artificially create semantic search terms every time a search is conducted. This is a burdensome task to create a satisfactory query. Often the user will have to redo the query more than once. The time spent on this task is quite burdensome and would include expensive on-line search time to stay on the commercial data base.

Using a list of words to represent the content of documents is a technique that also has problems of it's own. In this technique, the fact that words are ambiguous can cause documents to be retrieved that are not relevant to the search query. Further, relevant documents can exist that do not use the same words as those provided in the query. Using semantics addresses these concerns and can improve retrieval performance. Prior art has focussed on processes for disambiguation. In these processes, the various meanings of words (also referred to as senses) are pruned (reduced) with the hope that the remaining meanings of words will be the correct one. An example of well known pruning processes is U.S. Pat. No. 5,056,021 which is incorporated by reference.

However, the pruning processes used in disambiguation cause inherent problems of their own. For example, the correct common meaning may not be selected in these processes. Further, the problems become worse when two separate sequences of words are compared to each other to determine the similarity between the two. If each sequence is disambiguated, the correct common meaning between the two may get eliminated.

The inventor of the subject invention has used semantics to avoid the disambiguation problem. See U.S. patent application Ser. No. 08/148,688 filed on Nov. 5, 1993 which issued as U.S. Pat. No. 5,576,954 on Nov. 19, 1996. For semantics, the various meanings of words are not pruned but combined with the various meanings of other words and the statistically common meanings for small groups of words yield the correct common meaning for those words. This approach has been shown to improve the statistical ranking of retrieved information. In the semantic approach, the prunning process for common meaning is replaced by a statistical determination of common meaning. Crucial to this approach is the fact that retrieval documents must be small.

Relevance feedback has sometimes been used to improve statistical ranking. For relevance feedback, the judgements of the user concerning viewed information are used to automatically modify the search for more information. However, in relevance feedback, conventional IR (Information Retrieval) systems have a limited recall. G. Salton, Automatic Information Organization and Retrieval, McGraw-Hill, 1968. This limited recall causes only a few relevant documents are retrieved in response to user queries if the search process is based solely on the initial query. This limited recall indicates a need to modify (or reformulate) the initial query in order to improve performance. During this reformulation, it is customary to have to search the relevant documents iteratively as a sequence of partial search operations. The results of earlier searches can be used as feedback information to improve the results of later searches. One possible way to do this is to ask the user to make a relevance decision on a certain number of retrieved documents. Then this relevance information can be manually used to construct an improved query formulation and recalculate the similarities between documents and query in order to rank them. This process is known as relevance feedback.

A basic assumption behind relevance feedback is that, for a given query, documents relevant to it should resemble each other in a sense that they have reasonably similar keyword content. This implies that if a retrieved document is identified as relevant, then the initial query can be modified to increase its similarity to such a relevant document. As a result of this reformulation, it is expected that more of the relevant documents and fewer of the nonrelevant documents will be extracted. The automatic construction of an improved query is actually straightforward, but it does increase the complexity of the user interface and the use of the retrieval system, and it can slow down query response time. Essentially, document information viewed as relevant to a query can be used to modify the weights of terms and semantic categories in the original query. A modification can also be made using documents viewed as not relevant to a query.

The main problems with using relevance feedback are many. First, the original query becomes very large whenever all the words in a viewed relevant document are added to the original query. Secondly, it takes a long time to read large documents and decide if they are relevant or not. Another problem is that often only part of a large document is actually relevant. Other patents have tried to address this problem. See U.S. Pat. No. 5,297,027 to Morimoto et al.

The inventor is not aware of any prior art that combines statistical ranking, semantics, relevance feedback and using sentences (or clauses) as documents when queries are expressed in natural language in order to be able to search for and retrieve relevant documents.

SUMMARY OF THE INVENTION

The first objective of the present invention is to provide a natural language retrieval system which combines statistical ranking, semantics, relevance feedback and using sentences (or clauses) as documents when using natural language queries in order to be able to search for and retrieve relevant documents.

The second object of this invention is to provide an automated document retrieval system that minimizes the reading efforts of the user.

The third object of this invention is to provide an automated document retrieval system that minimizes the need for highlighting relevant words on a screenful of text in order to be able to indicate relevant information from a query.

The prefered method of the invention uses statistical ranking and the concept of semantics as shown in U.S. patent application Ser. No. 08/148,688 filed on Nov. 5, 1993, in order to rank relevant documents retrieved for a user's original query. After submitting a query, the user then reads one or more of the topmost documents in the ranked list of documents produced for the query. Since each document is very small (a clause, or a sentence at most), it is very easy for the user to quickly indicate if the document is relevant or not relevant to the original query. For each document flagged as relevant or not relevant, an automatic modification is made to the original query to essentially increase or decrease the importance of words. The new query is used to create another ranked list of documents. The feedback process repeats until the user stops the process.

In the subject invention, semantics helps to push relevant documents upward in a statistically ranked list. Relevance feedback helps the user automatically identify alternative words useful for expressing the query. The effort displayed by the user is minimal since the user views only small amounts of text and makes only a single judgement call on whether the small piece of text is relevant or not relevant for each small amount of text.

The invention can be applied to tasks such as retrieving documents relevant to a search request (sometimes referred to as archival retrieval), filtering documents which are relevant to a search request (sometimes referred to as routing) and answering questions from general information data bases.

Further objects and advantages of this invention will be apparent from the following detailed description of a presently preferred embodiment which is illustrated schematically in the accompanying drawings.

BRIEF DESCRIPTION OF THE FIGURES

FIG. 1 illustrates the preferred embodiment of the invention.

FIG. 2 illustrates the procedure used in patent application with Ser. No. 08/148,688 (filed on Nov. 5, 1993 which issued as U.S. Pat. No. 5,576,954 on Nov. 19, 1996) to determine a number to indicate the relevance or similarity of a document to a query.

FIG. 3 illustrates an example of an original user query and a collection of eight documents.

FIG. 4 is a list of words considered too general to have any value as a keyword, or as a word having any useful semantic value.

FIG. 5 is a list of words used in the original query of FIG. 3; this list becomes Query Word List in Step 100 of FIG. 1.

FIG. 6 provides the list of words used in each of the eight documents of FIG. 3.

FIG. 7 is a list of statistical data for all the words in the eight documents of FIG. 3; the information shown is a count of the number of documents containing each word, and the IDF of each word.

FIG. 8 reveals semantic information about each word used in the original query in FIG. 3; for each word listed in FIG. 5, this figure shows a count of the semantic categories triggered by the word, along with a list of the numeric codes for those categories. This information comes for Roget's International Thesaurus (5th Edition), edited by Robert L. Chapman, HarperCollins Publishers, 1992.

FIGS. 9A-9E reveal semantic information about each word used in the collection of eight documents in FIG. 3; for each word listed in FIG. 6, this figure show a count of the semantic categories triggered by the word, along with a list of the numeric codes for those categories. This information comes from Roget's International Thesaurus (5th Edition), edited by Robert L. Chapman, HarperCollins Publishers, 1992.

FIG. 10 provides the Document List of DocIds created in Step 200 of FIG. 1 for the example of FIG. 3.

FIG. 11 is a list of the eight documents in the example of FIG. 3 ranked in order of their relevance or similarity (SIM value) to the words used in the original query of FIG. 3 and shown in FIG. 5; both the DocId and the SIM value are shown as a pair in this list. This list is a sorted Relevancy List created at Step 900 in FIG. 1.

FIG. 12 is a list of words in a second query built from the original query after removing the words found in Document 5 (only the word "travel" was removed). This list is created by Step 1300 in FIG. 1.

FIG. 13 is a list of seven documents in the example of FIG. 3 (Document 5 has been removed) ranked in order of their relevance or similarity (SIM value) to the words of the second query of FIG. 12; both the DocID and the SIM value are shown as a pair in this list. This list is a sorted Relevancy List created at Step 900 in FIG. 1.

FIG. 14 is a list of words in a third query built by adding words found in Document 4 to the words of the second query of FIG. 12; this list is created by Step 1200 in FIG. 1.

FIG. 15 is a list of six documents in the example of FIG. 3 (Document 5 and Document 4 have been removed) ranked in order of their relevance or similarity (SIM value) to the words of the third query of FIG. 14; both the DocId and the SIM value are shown as a pair in this list. This list is a sorted Relevancy List created by Step 900 in FIG. 1. The top document on this list (Document 2) provides the answer to the original query of FIG. 3.

DESCRIPTION OF THE PREFERRED EMBODIMENT

Before explaining the disclosed embodiment of the present invention in detail it is to be understood that the invention is not limited in its application to the details of the particular arrangement shown since the invention is capable of other embodiments. Also, the terminology used herein is for the purpose of description and not of limitation.

A prototype of the inventor's process has been successfully used at the NASA KSC Public Affairs Office. The performance of the prototype was measured by a count of the number of documents one must read in order to find an answer to a natural language question. In some queries, a noticeable semantic improvement has been observed. For example, if only keywords are used for the query "How fast does the orbiter travel on orbit?" then 17 retrieved paragraphs must be read to find the answer to the query. But if semantic information is used in conjunction with key words then only 4 retrieved paragraphs need to be read to find the answer to the query. Thus, the prototype enabled a searcher to find the answer to their query by a substantial reduction of the number of documents that must be read.

Reference will now be made in detail to the present preferred embodiment of the invention as illustrated in the accompanying drawings.

The present preferred embodiment is demonstrated using an environment where a user's original query is a simple question and the user is searching for an answer to the question. During the search, we expect the user to see relevant and non-relevant documents. The user is expected to continue until a document answering the question is read, or until there are no more documents left to read.

The detailed description refers to acronyms and terminology that is described in the following chart.

TERMINOLOGY

______________________________________SIM for a query     A number which measures the relevance of a docu-and a document     ment to a query.qword     A word in the list of words used in a query.cat       A semantic category code.qp        The probability a qword triggers a cat.dword     A word in the list of words used in a document.dp        The probability a dword triggers a cat.DocId     The identifier for a document, the document numberN         Total number of documents.NDOCS     The number of documents a word is in.for a wordIDF       The inverse document frequency which is definedfor a word     here to be log2 (N/NDOCS for the word).Document  List of words used in a document. FIG. 6 showsWord List eight of these lists.Query     List of words used in a query.Word ListDocument List     List of DocIds.Relevancy List     List of DocId, SIM pairs.______________________________________
Statistical Ranking with Semantics

FIG. 2 illustrates the procedure used in U.S. patent application Ser. No. 08/148,688 (filed on Nov. 5, 1993 which issued as U.S. Pat. No. 5,576,954 on Nov. 19, 1996) to determine a number to indicate the relevance or similarity of a document to a query. The procedure is based on the existence of a semantic lexicon. For a given word, the semantic lexicon indicates all the senses (different meanings) of a word. Roget's International Thesaurus (5th Edition), edited by Robert L. Chapman, HarperCollins Publishers, 1992 can be used as a semantic lexicon. The procedure illustrated in FIG. 2 also uses a statistical similarity calculation.

To illustrate, FIG. 3 provides an original user query (a question) and a collection of eight documents, where each document is a sentence and has a DocId which is an integer number. Notice that Document 2 explicitly answers the user query.

In statistical systems it is common to have a list of words which can be ignored because they are relatively useless as keywords. FIG. 4 provides a list of words not used for this example. Using the list of words not used, the example of FIG. 3 can be transformed into the words used in the original query of FIG. 5 and the words used in each of the eight documents of FIG. 6.

FIG. 7 provides a list of statistical data for the words used in all of the eight documents, in alphabetical order. The number of documents that each word is in is shown in the second column of the table. This is called NDOCS for a word. The third column of the table in FIG. 7 indicates a measure of the importance of each word.

The formula used for calculating the importance of a word is a statistical formula. A good one to use for this example is the inverse document frequency (IDF) formula:

IDF of a word=log2 (N/NDOCS for the word)

where N is the total number of documents (8) and NDOCS is the number of documents a word is in. For example, since "orbit" is in 4 documents,

IDF for orbit=log2 (8/4)=log2 (2)=1

and since "increase" is in one document,

IDF for increase=log2 (8/1)=log2 (8)=3.

These IDF numbers are recorded in the third column of FIG. 7. It is clear that words which are in many documents are less important (as search words) than words which are in only a few documents.

FIG. 8 provides the semantics of the words in the original query. For each word used in the original query (FIG. 5), the second column shows the number of senses (meanings) the word has in Roget's Thesaurus, and the third column lists the numeric codes for those different meanings.

FIG. 9 provides the semantics of the words used in the eight documents. For each word used in the eight documents (FIG. 7), the second column shows the number of senses (meanings) the word has in Roget's Thesaurus, and the third column lists the numeric codes for those different meanings.

Notice that all but one of the words used in the query are used in the eight documents. The word "fast" does not appear in the eight documents.

For this example, a semantic category will be a "large category" in Roget's Thesaurus. There are 1073 large categories. The number of smaller categories will be used to determine a probability for a specific large category. For example, consider the word "fast", which triggers category "174.15" and category "174.17"; each of these is in the large category "174". So, the word "fast" triggers category "174" with a probability of 2/15 since 15 is the number of smaller categories triggered by the word "fast".

Also in this example, the weight of a word in a document will be the frequency of the word in the document multiplied by the word's IDF value. In the example, all frequencies turn out to be 1, so the weight of a word in a document becomes the word's IDF value.

The calculation of a SIM value for a query and a document can now be explained by reference to the Similarity Procedure in FIG. 2 and a small sample calculation. Consider the words used in the original query of FIG. 5 and the words used in Document 4 of FIG. 6. These two lists are called the Query Word List and the Document Word List, and they are the inputs to the Similarity Procedure.

Step 405 sets the SIM value to zero. Step 410 sets qword to "fast". Since "fast" is not in Document 4, Step 420 causes movement to step 430. Since "fast" does trigger semantic categories, Step 430 causes movement to Step 435 and Step 440 causes cat to be "515" and qp to be 1/15. At Step 445, there is no word in Document 4 that triggers "515" so Step 435 is executed again. Steps 435, 440, and 445 re repeatedly executed with no movement to Step 450 until category "174" is used. At Step 440, cat eventually becomes "174" and qp becomes 2/15 since there are two of "174" in the list of categories triggered by "fast". At Step 450, dword becomes "velocity" since "velocity" triggers "174". Also, dp becomes 1/3 since "velocity" triggers three separate categories.

At Step 455, notice that since "fast" is not a word in any of the documents, its IDF is not defined in FIG. 7; so, in this case, the IDF of the word "velocity" is substituted. Another possibility in this case is to substitute a very high IDF value for undefined IDF values. At Step 455, SIM is increased by

(2/15*1)*(1/3*1)=0.0444

so SIM now equals 0.0444.

Eventually, at Step 435, there are no more categories triggered by "fast" and this causes movement to Step 410.

At Step 410, "orbit" is the next word in the query and, at Step 415, qword now becomes "orbit". At Step 420, the fact that "orbit" is also in Document 4 causes movement to Step 425. At Step 425, SIM is increase by the weight of "orbit" in the query multiplied by the weight of "orbit" in Document 4, and this amount is

(1)*(1)=1.0000

so SIM now equals 1.0444.

At Step 430, since "orbit" also triggers semantic categories, there is movement to Step 435. Steps 435, 440, and 445 are repeatedly executed for the semantic categories triggered by "orbit". For category "245" triggered by "orbit", the word "increase" in Document 4 is also a trigger. So, when cat becomes "245" and qp becomes 1/13, Step 450 causes dword to become "increase" and dp to become 1/20. Then, at Step 455, SIM is increased by

(1/13*1)*(1/20*3)=0.0154

so SIM now equals 1.0598. Note that the IDF of "increase" is 3, and so the weight of "increase" in Document 4 is 3.

Notice that Step 445 does not select the word "orbit" in Document 4, since qword is "orbit" and the semantic contribution of "orbit" in Document 4 was handled earlier by Step 425. Eventually, at Step 435, there are no more categories triggered by "orbit" and this causes movement to Step 410.

At Step 410, "orbiter" is the next word in the query and at Step 415, qword now becomes "orbiter". Since "orbiter" is also in Document 4, Step 420 causes movement to Step 425. At Step 425, SIM is increased by the weight of "orbiter" in the query multiplied by the weight of "orbiter" in Document 4, and this amount is

(1)*(1)-1.0000

so SIM now equals 2.0598.

At Step 430, since "orbiter" does not trigger any semantic categories, there is movement to Step 410.

At Step 410, "travel" is the next (and last) word in the query and, at Step 415, qword now becomes "travel". Since "travel" is not in Document 4, Step 420 causes movement to Step 430. Since "travel" does trigger semantic categories, Step 430 causes movement to Step 435 and Step 440 causes cat to be "162" and qp to be 2/9 since "travel" triggers "162.1" and "162.2". At Step 445, there is no word in Document 4 that triggers "162", so Step 435 is executed again. Steps 435, 440, and 445 are repeatedly executed with no movement to Step 450 until category "172" is used, and category "177" is used.

When Step 440 causes cat to become "172" and qp to be 2/9, Step 445 causes movement to Step 450. The value of qp is 2/9 because "travel" triggers "172.2" and "172.5". At Step 450, dword becomes "velocity" and dp becomes 1/3 since "velocity" triggers "172" among three triggered separate categories. At Step 455, SIM is increased by

(2/9*3)*(1/3*1)=0.2222

so SIM now equals 2.2820.

When Step 440 causes cat be become "177" and qp to be 4/9, Step 445 causes movement to Step 450. The value of qp is 4/9 because "travel" triggers "177", "177.1", "177.18", and "177.21". At Step 450, dword becomes "velocity" and dp becomes 1/3. At Step 455, SIM is increased by

(4/9*3)*(1/3*1)=0.4444

so SIM now equals 2.7264.

Eventually, at Step 435, there are no more categories triggered by "travel" and this causes movement to Step 410. At Step 410, the procedure for calculating SIM stops because there are no more words in the query.

The final value of SIM is 2.7264 and this represents a measure of the similarity between the original query in FIG. 3 and Document 4 in FIG. 3. The DocId of 4 and the SIM value of 2.7264 are the outputs of the Similarity Procedure.

Relevance Feedback with Small Amounts of Text

FIG. 1 illustrates the preferred embodiment of the invention. The Feedback Procedure of FIG. 1 activates the Similarity Procedure of FIG. 2 many times. To illustrate, FIG. 3 provides an original user query (a question) and a collection of eight documents, where each document is a sentence and has a DocId which is an integer. Notice that Document 2 explicitly answers the user query.

This is a question/answer environment and the preferred embodiment of the invention is designed for this environment. The invention will help the user retrieve Document 2 (the answer to the user query in FIG. 3).

At Step 100, Query Word List is set to the list of four words used in the original user query and shown in FIG. 5. At Step 200, Document Word List is set to the list of eight DocIds shown in FIG. 10. At Step 300, Relevancy List is set to be empty. Eventually, Relevancy List will be a list of DocId, SIM pairs sorted by SIM value to represent a ranking of the documents based on their statistical similarity to the query.

At Step 400, DocId is set equal to the first document identifier in Document List. DocId is set to Document 1.

At Step 500, the Query Word List of FIG. 5 and the Document Word List for Document 1 in FIG. 6 are input to the Similarity Procedure of FIG. 2. The output of the Similarity Procedure is DocId of 1 and SIM of 2.0338.

At Step 600, the pair DocId of 1 and SIM of 2.0338 is added to the Relevancy List. Since there are more DocIds to process in Document List, Step 700 causes movement to Step 800 where DocId becomes Document 2. Then Step 500 activates the Similarity Procedure, again. Steps 500, 600, 700, and 800 cause the Similarity Procedure to be activated for each DocId in Document List, along with addition of the DocIds and their SIM values as pairs in Relevancy List. Eventually, Step 700 causes movement to Step 900 where the Relevancy List is sorted on SIM value.

FIG. 11 reveals the result of Step 900 for the original user query and the eight documents of FIG. 3. Statistical keyword and semantic ranking has determined that Document 5 is the most relevant document for the original user query, Document 4 is the next most relevant document for the original query, and so on.

At Step 1000, DocId is set to Document 5 and the document

"Atlantis will travel more than half a million miles in ocean research."

is shown to the user at Step 1100 where the user must decide if the sentence is relevant, not relevant, or answers the original query. The sentence is obviously not relevant, so Step 1100 causes movement to Step 1300. At Step 1300, any word in the Document Word List for Document 5 (as shown in FIG. 6) is removed from the Query Word List of FIG. 5; the result is shown in FIG. 12 where the word "travel" has been removed. The Query Word List now has three words in it, and it becomes the automatically built second query.

At Step 1400, DocId of 5 is removed from the Document List since the user has read the document. Since there are still seven documents in Document List, Step 1500 causes movement to Step 300 where the Relevancy List is set to empty, again.

At Step 400, DocId is set equal to Document 1 again and Steps 500, 600, 700, and 800 cause the activation of the Similarity Procedure of FIG. 2 for computing the similarity of the second query to each of the remaining seven documents, along with addition of the DocIds and their SIM values in Relevancy List. Eventually, Step 700 causes movement to Step 900 where the Relevancy List is sorted on SIM value.

FIG. 13 reveals the result of Step 900 for the second query and the seven documents not read by the user. Statistical keyword and semantic ranking has determined that Document 4 is now the most relevant document.

At Step 1000, DocId is set to Document 4 and the document

"The engines are used to increase the velocity of the orbiter on orbit."

is shown to the user at Step 1100 where the user must decide if the sentence is relevant, not relevant, or answers the original query. Most people would agree that the sentence is relevant, so Step 1100 causes movement to Step 1200.

At Step 1200, the words in the Document Word List for Document 4 (as shown in FIG. 6) are added to the Query Word List for the second query of FIG. 12; the result is shown in FIG. 14 where the words "engines", "increase", and "velocity" are added. The Query Word List now has six words in it, and it becomes the automatically built third query.

At Step 1400, DocId of 4 is removed from the Document List since the user has read the document. Since there are still six documents in the Document List, Step 1500 causes movement to Step 300 where the Relevancy List is set to empty, again.

At Step 400, DocId is set equal to Document 1 again and Steps 500, 600, 700, and 800 cause the activation of the Similarity Procedure of FIG. 2 for computing the similarity of the third query to each of the remaining six documents, along with addition of the DocIds and their SIM values in Relevancy List. Eventually, Step 700 causes movement to Step 900 where the Relevancy List is sorted on SIM value.

FIG. 15 reveals the result of Step 900 for the third query and the six documents not yet read by the user. Statistical keyword and semantic ranking has determined that Document 2 is now the most relevant document.

At Step 1000, DocId is set to Document 2 and the document

"The orbiter's engines maintain a velocity on orbit of approximately 25,405 feet per second."

is shown to the user at Step 1100 where the user must decide if the sentence is relevant, not relevant, or answers the original query. Obviously, Document 2 provides the answer to the original query, so the retrieval process stops after three sentences were read.

The feedback and sentencer features are quite useful to user in saving time and enhancing the quality of the search. The feedback feature of the subject invention helps to introduce new words and gets rid of bad words, e.g. the word travel is removed from FIG. 5 and "velocity" is added in FIG. 14.

The sentencer minimizes reading time and allows the user to make their relevancy decisions very easy by just requiring the user to indicate by a key stroke whether a document is relative or not relative. In addition, the sentencer saves the user time by forcing the user to discover small "units" which are relevant or not relevant and the decision is easy. While the prefered embodiment has been described in reference to one type of document collection, the invention can be equally applicable to all types of documents such as but not limited to patents, legal documents, medical documents, articles, journals and the like.

Further, there is no size limit to the number of documents that can be searched.

The invention can be incorporated on personal computers to search for internal files and can be applied to modem search systems accessible to DIALOG, ORBIT, and the like.

While the invention has been described, disclosed, illustrated and shown in various terms of certain embodiments or modifications which it has presumed in practice, the scope of the invention is not intended to be, nor should it be deemed to be, limited thereby and such other modifications or embodiments as may be suggested by the teachings herein are particularly reserved especially as they fall within the breadth and scope of the claims here appended.

Patent Citations
Cited PatentFiling datePublication dateApplicantTitle
US5251316 *Jun 28, 1991Oct 5, 1993Digital Equipment CorporationMethod and apparatus for integrating a dynamic lexicon into a full-text information retrieval system
US5293552 *Mar 30, 1992Mar 8, 1994U.S. Philips CorporationMethod for storing bibliometric information on items from a finite source of text, and in particular document postings for use in a full-text document retrieval system
US5297027 *May 6, 1991Mar 22, 1994Hitachi, Ltd.Method of and apparatus for promoting the understanding of a text by using an abstract of that text
US5297039 *Jan 27, 1992Mar 22, 1994Mitsubishi Denki Kabushiki KaishaText search system for locating on the basis of keyword matching and keyword relationship matching
US5299125 *Feb 8, 1993Mar 29, 1994Semantic Compaction SystemsNatural language processing system and method for parsing a plurality of input symbol sequences into syntactically or pragmatically correct word messages
US5301109 *Jul 17, 1991Apr 5, 1994Bell Communications Research, Inc.Computerized cross-language document retrieval using latent semantic indexing
US5305208 *Aug 6, 1991Apr 19, 1994Kabushiki Kaisha ToshibaDatabase retrieval system for effectively displaying differences between elements of plural drawings
US5309359 *Aug 16, 1990May 3, 1994Boris KatzMethod and apparatus for generating and utlizing annotations to facilitate computer text retrieval
US5319790 *Jun 12, 1992Jun 7, 1994Canon Kabushiki KaishaApparatus automatically entering mode for storing document and mode for reading document in response to representative data indicating whether stored document exists corresponding to input title
US5321608 *Nov 27, 1991Jun 14, 1994Hitachi, Ltd.Method and system for processing natural language
US5321609 *May 27, 1992Jun 14, 1994Franklin Electronic Publishers, IncorporatedElectronic encyclopedia
US5321833 *Aug 29, 1990Jun 14, 1994Gte Laboratories IncorporatedAdaptive ranking system for information retrieval
US5323313 *Apr 25, 1991Jun 21, 1994Analytic Legal ProgramsDesign and implementation of drafting possible permutations of a set of text
US5325298 *Sep 3, 1991Jun 28, 1994Hnc, Inc.Methods for generating or revising context vectors for a plurality of word stems
US5325465 *Mar 4, 1992Jun 28, 1994Singapore Computer Systems LimitedEnd user query facility
US5331554 *Dec 10, 1992Jul 19, 1994Ricoh CorporationMethod and apparatus for semantic pattern matching for text retrieval
US5331556 *Jun 28, 1993Jul 19, 1994General Electric CompanyMethod for natural language data processing using morphological and part-of-speech information
US5351190 *Oct 10, 1991Sep 27, 1994Canon Kabushiki KaishaCommunication apparatus and method establishing a language dependent sort order search
US5404295 *Jan 4, 1994Apr 4, 1995Katz; BorisMethod and apparatus for utilizing annotations to facilitate computer retrieval of database material
US5418951 *Sep 30, 1994May 23, 1995The United States Of America As Represented By The Director Of National Security AgencyMethod of retrieving documents that concern the same topic
US5537586 *May 6, 1994Jul 16, 1996Individual, Inc.Enhanced apparatus and methods for retrieving and selecting profiled textural information records from a database of defined category structures
US5544352 *Jun 14, 1993Aug 6, 1996Libertech, Inc.Method and apparatus for indexing, searching and displaying data
US5598557 *Sep 22, 1992Jan 28, 1997Caere CorporationApparatus and method for retrieving and grouping images representing text files based on the relevance of key words extracted from a selected file to the text files
US5642502 *Dec 6, 1994Jun 24, 1997University Of Central FloridaMethod and system for searching for relevant documents from a text database collection, using statistical ranking, relevancy feedback and small pieces of text
US5694592 *Aug 28, 1995Dec 2, 1997University Of Central FloridaProcess for determination of text relevancy
US5724567 *Apr 25, 1994Mar 3, 1998Apple Computer, Inc.Information access system
Referenced by
Citing PatentFiling datePublication dateApplicantTitle
US5987460 *Jul 3, 1997Nov 16, 1999Hitachi, Ltd.Document retrieval-assisting method and system for the same and document retrieval service using the same with document frequency and term frequency
US6088692 *Apr 5, 1999Jul 11, 2000University Of Central FloridaNatural language method and system for searching for and ranking relevant documents from a computer database
US6219664 *Aug 27, 1999Apr 17, 2001International Business Machines Corp.Search method and system using syntactic information
US6292796 *Feb 23, 1999Sep 18, 2001Clinical Focus, Inc.Method and apparatus for improving access to literature
US6363378 *Oct 13, 1998Mar 26, 2002Oracle CorporationRanking of query feedback terms in an information retrieval system
US6370525 *Nov 13, 2000Apr 9, 2002Kcsl, Inc.Method and system for retrieving relevant documents from a database
US6385611 *May 5, 2000May 7, 2002Carlos CardonaSystem and method for database retrieval, indexing and statistical analysis
US6493692Sep 30, 1999Dec 10, 2002Canon Kabushiki KaishaInformation search apparatus and method, and computer readable memory
US6493705 *Sep 30, 1999Dec 10, 2002Canon Kabushiki KaishaInformation search apparatus and method, and computer readable memory
US6505196 *Aug 14, 2001Jan 7, 2003Clinical Focus, Inc.Method and apparatus for improving access to literature
US6526400Sep 30, 1999Feb 25, 2003Canon Kabushiki KaishaInformation search apparatus and method
US6738764 *May 8, 2001May 18, 2004Verity, Inc.Apparatus and method for adaptively ranking search results
US6766316Jan 18, 2001Jul 20, 2004Science Applications International CorporationMethod and system of ranking and clustering for document indexing and retrieval
US6804662Oct 27, 2000Oct 12, 2004Plumtree Software, Inc.Method and apparatus for query and analysis
US6826724 *Dec 27, 1999Nov 30, 2004Ricoh Company, Ltd.Document processor, document classification device, document processing method, document classification method, and computer-readable recording medium for recording programs for executing the methods on a computer
US6829599Oct 2, 2002Dec 7, 2004Xerox CorporationSystem and method for improving answer relevance in meta-search engines
US6842876 *Apr 14, 1998Jan 11, 2005Fuji Xerox Co., Ltd.Document cache replacement policy for automatically generating groups of documents based on similarity of content
US6850937 *Aug 22, 2000Feb 1, 2005Hitachi, Ltd.Word importance calculation method, document retrieving interface, word dictionary making method
US6856987 *Sep 5, 2002Feb 15, 2005Canon Kabushiki KaishaInformation search apparatus and method, and computer readable memory
US6999974 *Feb 12, 2003Feb 14, 2006International Business Machines CorporationMethod and system for improving a text search
US7003511 *Aug 2, 2002Feb 21, 2006Infotame CorporationMining and characterization of data
US7027974Oct 27, 2000Apr 11, 2006Science Applications International CorporationOntology-based parser for natural language processing
US7047482 *Feb 28, 2001May 16, 2006Gary OdomAutomatic directory supplementation
US7054860Oct 6, 2003May 30, 2006Hitachi, Ltd.Method and system for retrieving a document and computer readable storage medium
US7054861Oct 13, 2004May 30, 2006Canon Kabushiki KaishaInformation search apparatus and method, and computer readable memory
US7054882 *Feb 12, 2003May 30, 2006International Business Machines CorporationMethod and system for improving a text search
US7072888 *Jun 16, 1999Jul 4, 2006Triogo, Inc.Process for improving search engine efficiency using feedback
US7080068Jun 27, 2001Jul 18, 2006Thomas LeitermannAutomatic search method
US7146381 *Dec 20, 1999Dec 5, 2006Actioneer, Inc.Information organization and collaboration tool for processing notes and action requests in computer systems
US7155334Sep 29, 2005Dec 26, 2006Honeywell International Inc.Use of sensors in a state observer for a diesel engine
US7165399Dec 29, 2004Jan 23, 2007Honeywell International Inc.Method and system for using a measure of fueling rate in the air side control of an engine
US7182075Dec 7, 2004Feb 27, 2007Honeywell International Inc.EGR system
US7188106Apr 30, 2002Mar 6, 2007International Business Machines CorporationSystem and method for aggregating ranking results from various sources to improve the results of web searching
US7216073Mar 13, 2002May 8, 2007Intelligate, Ltd.Dynamic natural language understanding
US7275374Mar 30, 2005Oct 2, 2007Honeywell International Inc.Coordinated multivariable control of fuel and air in engines
US7295965Jun 29, 2001Nov 13, 2007Honeywell International Inc.Method and apparatus for determining a measure of similarity between natural language sentences
US7328208Nov 9, 2005Feb 5, 2008International Business Machines CorporationMethod and system for improving a text search
US7328577Dec 29, 2004Feb 12, 2008Honeywell International Inc.Multivariable control for an engine
US7346608Sep 20, 2004Mar 18, 2008Bea Systems, Inc.Method and apparatus for query and analysis
US7357125Oct 26, 2005Apr 15, 2008Honeywell International Inc.Exhaust gas recirculation system
US7389773Aug 18, 2005Jun 24, 2008Honeywell International Inc.Emissions sensors for fuel control in engines
US7403938Sep 20, 2002Jul 22, 2008Iac Search & Media, Inc.Natural language query processing
US7415389Dec 29, 2005Aug 19, 2008Honeywell International Inc.Calibration of engine control systems
US7467614Dec 29, 2004Dec 23, 2008Honeywell International Inc.Pedal position and/or pedal change rate for use in control of an engine
US7469177Jun 17, 2005Dec 23, 2008Honeywell International Inc.Distributed control architecture for powertrains
US7483877 *Apr 11, 2003Jan 27, 2009International Business Machines CorporationDynamic comparison of search systems in a controlled environment
US7496561Dec 1, 2003Feb 24, 2009Science Applications International CorporationMethod and system of ranking and clustering for document indexing and retrieval
US7584221Mar 18, 2004Sep 1, 2009Microsoft CorporationField weighting in text searching
US7591135Dec 28, 2006Sep 22, 2009Honeywell International Inc.Method and system for using a measure of fueling rate in the air side control of an engine
US7599917Aug 15, 2005Oct 6, 2009Microsoft CorporationRanking search results using biased click distance
US7603616Nov 5, 2004Oct 13, 2009Microsoft CorporationProxy server using a statistical model
US7664803Jan 11, 2006Feb 16, 2010Canon Kabushiki KaishaInformation search apparatus and method, and computer readable memory
US7707201 *Dec 5, 2005Apr 27, 2010Yahoo! Inc.Systems and methods for managing and using multiple concept networks for assisted search processing
US7716198 *Dec 21, 2004May 11, 2010Microsoft CorporationRanking search results using feature extraction
US7739277Sep 30, 2004Jun 15, 2010Microsoft CorporationSystem and method for incorporating anchor text into ranking search results
US7743061Nov 12, 2003Jun 22, 2010Proximate Technologies, LlcDocument search method with interactively employed distance graphics display
US7743606Nov 18, 2004Jun 29, 2010Honeywell International Inc.Exhaust catalyst system
US7752840Mar 24, 2005Jul 13, 2010Honeywell International Inc.Engine exhaust heat exchanger
US7761448Sep 30, 2004Jul 20, 2010Microsoft CorporationSystem and method for ranking search results using click distance
US7765792Oct 21, 2005Aug 3, 2010Honeywell International Inc.System for particulate matter sensor signal processing
US7792833Apr 26, 2006Sep 7, 2010Microsoft CorporationRanking search results using language types
US7827181Sep 29, 2005Nov 2, 2010Microsoft CorporationClick distance determination
US7840400Nov 21, 2006Nov 23, 2010Intelligate, Ltd.Dynamic natural language understanding
US7840569Oct 18, 2007Nov 23, 2010Microsoft CorporationEnterprise relevancy ranking using a neural network
US7854009Jun 12, 2003Dec 14, 2010International Business Machines CorporationMethod of securing access to IP LANs
US7856350Aug 11, 2006Dec 21, 2010Microsoft CorporationReranking QA answers using language modeling
US7878178Jun 23, 2008Feb 1, 2011Honeywell International Inc.Emissions sensors for fuel control in engines
US7917497Apr 18, 2008Mar 29, 2011Iac Search & Media, Inc.Natural language query processing
US7941446Dec 18, 2009May 10, 2011Xerox CorporationSystem with user directed enrichment
US7949581Sep 7, 2006May 24, 2011Patentratings, LlcMethod of determining an obsolescence rate of a technology
US7962511Apr 29, 2003Jun 14, 2011Patentratings, LlcMethod and system for rating patents and other intangible assets
US7974974Mar 20, 2008Jul 5, 2011Microsoft CorporationTechniques to perform relative ranking for search results
US8014997Sep 20, 2003Sep 6, 2011International Business Machines CorporationMethod of search content enhancement
US8027966Aug 21, 2008Sep 27, 2011International Business Machines CorporationMethod and system for searching a multi-lingual database
US8027994Aug 21, 2008Sep 27, 2011International Business Machines CorporationSearching a multi-lingual database
US8082246Jul 1, 2010Dec 20, 2011Microsoft CorporationSystem and method for ranking search results using click distance
US8095565May 5, 2006Jan 10, 2012Microsoft CorporationMetadata driven user interface
US8109255Dec 20, 2010Feb 7, 2012Honeywell International Inc.Engine controller
US8156125Feb 19, 2008Apr 10, 2012Oracle International CorporationMethod and apparatus for query and analysis
US8161036Jun 27, 2008Apr 17, 2012Microsoft CorporationIndex optimization for ranking using a linear model
US8165786Jul 23, 2010Apr 24, 2012Honeywell International Inc.System for particulate matter sensor signal processing
US8171031Jan 19, 2010May 1, 2012Microsoft CorporationIndex optimization for ranking using a linear model
US8175989Jan 3, 2008May 8, 2012Choicestream, Inc.Music recommendation system using a personalized choice set
US8219557Jun 10, 2010Jul 10, 2012Xerox CorporationSystem for automatically generating queries
US8239216 *Jan 9, 2009Aug 7, 2012Cerner Innovation, Inc.Searching an electronic medical record
US8239413Apr 1, 2011Aug 7, 2012Xerox CorporationSystem with user directed enrichment
US8265854Jul 8, 2011Sep 11, 2012Honeywell International Inc.Configurable automotive controller
US8266144Jul 1, 2011Sep 11, 2012Microsoft CorporationTechniques to perform relative ranking for search results
US8352511Aug 28, 2008Jan 8, 2013Partnet, Inc.Systems and methods for providing a confidence-based ranking algorithm
US8360040Jan 18, 2012Jan 29, 2013Honeywell International Inc.Engine controller
US8380731 *Dec 13, 2007Feb 19, 2013The Boeing CompanyMethods and apparatus using sets of semantically similar words for text classification
US8504175Jun 2, 2010Aug 6, 2013Honeywell International Inc.Using model predictive control to optimize variable trajectories and system control
US8543577Mar 2, 2011Sep 24, 2013Google Inc.Cross-channel clusters of information
US8620461Sep 24, 2009Dec 31, 2013Honeywell International, Inc.Method and system for updating tuning parameters of a controller
US8631009May 16, 2011Jan 14, 2014Steven LisaSystems and methods for embedded internet searching, and result display
US8719244 *Mar 23, 2005May 6, 2014Google Inc.Methods and systems for retrieval of information items and associated sentence fragments
US8725729Apr 3, 2006May 13, 2014Steven G. LisaSystem, methods and applications for embedded internet searching and result display
US8738635Jun 1, 2010May 27, 2014Microsoft CorporationDetection of junk in search result ranking
US8768700 *Sep 14, 2012Jul 1, 2014Google Inc.Voice search engine interface for scoring search hypotheses
US8793706Dec 16, 2010Jul 29, 2014Microsoft CorporationMetadata-based eventing supporting operations on data
US8812493Apr 11, 2008Aug 19, 2014Microsoft CorporationSearch results ranking using editing distance and document information
US8843486Sep 29, 2009Sep 23, 2014Microsoft CorporationSystem and method for scoping searches using index keys
US20100179827 *Jan 9, 2009Jul 15, 2010Cerner Innovation, Inc.Searching an electronic medical record
US20120290328 *Jul 24, 2012Nov 15, 2012Cerner Innovation, Inc.Searching an electronic medical record
USRE44452Dec 22, 2010Aug 27, 2013Honeywell International Inc.Pedal position and/or pedal change rate for use in control of an engine
EP1098258A1 *Nov 3, 1999May 9, 2001BRITISH TELECOMMUNICATIONS public limited companyInformation access
WO2000068757A2 *May 5, 2000Nov 16, 2000Carlos CardonaSystem and method for database retrieval, indexing and statistical analysis
WO2001033417A1 *Oct 20, 2000May 10, 2001British TelecommInformation access
WO2001042988A2 *Nov 7, 2000Jun 14, 2001Transcom Software IncComputer network information management system and method
WO2001056388A2 *Feb 5, 2001Aug 9, 2001Gen Dynamics Inf Systems IncAnnotating semantic ontologies
WO2001061555A2 *Feb 8, 2001Aug 23, 2001Martin PorterSearching station accessed by selection terminals
WO2001073599A2 *Mar 21, 2001Oct 4, 2001Koninkl Philips Electronics NvUser interface providing automatic generation and ergonomic presentation of keyword search criteria
WO2002001407A2 *Jun 27, 2001Jan 3, 2002Guru Netservices GmbhAutomatic search method
WO2003060790A1 *Dec 17, 2002Jul 24, 2003Joel C DonSystem and method of distributing public relations and marketing content
WO2004017178A2Aug 19, 2003Feb 26, 2004ChoicestreamStatistical personalized recommendation system
WO2007052883A1 *Aug 9, 2006May 10, 2007Jeong-Jin KimMethod for searching patent document by applying degree of similarity and system thereof
Classifications
U.S. Classification1/1, 707/E17.079, 707/E17.078, 707/E17.066, 707/E17.068, 715/234, 704/9, 707/E17.071, 707/999.005, 707/999.003
International ClassificationG06F17/30
Cooperative ClassificationY10S707/99939, Y10S707/99935, Y10S707/99933, Y10S707/99936, G06F17/30663, G06F17/30684, G06F17/30687, G06F17/30654, G06F17/3064
European ClassificationG06F17/30T2P4N, G06F17/30T2P4P, G06F17/30T2F1, G06F17/30T2F4, G06F17/30T2P2E
Legal Events
DateCodeEventDescription
Feb 28, 2011SULPSurcharge for late payment
Sep 22, 2010FPAYFee payment
Year of fee payment: 12
Dec 13, 2007ASAssignment
Owner name: UNIVERSITY OF CENTRAL FLORIDA RESEARCH FOUNDATION,
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:UNIVERSITY OF CENTRAL FLORIDA;REEL/FRAME:020234/0335
Effective date: 20071210
May 11, 2006FPAYFee payment
Year of fee payment: 8
Apr 16, 2002FPAYFee payment
Year of fee payment: 4
Jun 23, 1997ASAssignment
Owner name: CENTRAL FLORIDA, UNIVERSITY OF, FLORIDA
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:DRISCOLL, JAMES R.;REEL/FRAME:008662/0141
Effective date: 19941205